Performance of an automated registration-based method for OPENACCESS longitudinal lesion matching and comparison to inter-reader variability - Abstract

Q: What was the objective of the study?

To assess how well an automated lesion-matching algorithm performs compared with inter-reader variability when matching longitudinal lesions between scans in metastatic cancer patients.

Q: How were the scans and cohorts organized?

Forty paired longitudinal PET/CT and CT scan pairs were grouped into four cancer cohorts and further stratified by cancer burden into low, intermediate, and high groups.

Q: What did the results show about accuracy and efficiency?

Automated matching did not differ significantly from inter-reader variability across metrics in any cohort, and it was significantly more efficient than manual reader matching.

Sep 24, 2026

Uploaded by Bintang

Objective. Patients with metastatic disease are followed throughout treatment with medical imaging, and accurately assessing changes of individual lesions is critical for clinical decisions. The study evaluates an automated lesion-matching algorithm against inter-reader variability (IRV) in matching lesions across scans. Approach. Forty paired longitudinal PET/CT and CT datasets were grouped into four cancer cohorts and by burden (low, intermediate, high). Two nuclear medicine physicians independently matched lesions; reader agreement consensus served as gold standard. Results. Automated performance did not differ significantly from IRV for any metric in any cohort (p > 0.05). Automated matching was significantly more efficient (p < 0.001) while meeting IRV benchmarks. Conclusion. Automated lesion matching expedites analysis and improves longitudinal consistency.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Introduction

[https:](https://doi.org/10.1088/1361-6560/acef8f)[//](https://doi.org/10.1088/1361-6560/acef8f)[doi.org](https://doi.org/10.1088/1361-6560/acef8f)[/](https://doi.org/10.1088/1361-6560/acef8f)[10.1088](https://doi.org/10.1088/1361-6560/acef8f)[/](https://doi.org/10.1088/1361-6560/acef8f)[1361-6560](https://doi.org/10.1088/1361-6560/acef8f)[/](https://doi.org/10.1088/1361-6560/acef8f)[acef8f](https://doi.org/10.1088/1361-6560/acef8f) PAPER Performance of an automated registration-based method for OPENACCESS longitudinal lesion matching and comparison to inter-reader R5ECEIVMayE023 variability REVISED 25July2023 Daniel T Huff1,∗ , Victor Santoro-Fernandes2 , Song Chen3, Meijie Chen3, Carl Kashuk1, CEPTAuEgD FORust20U3BLICATION Amy J Weisman 1 , Robert Jeraj2,4 and Timothy G Perk1 1 AIQ Solutions, Madison, WI, United States ofAmerica PUBLISHED 28August2023 2 University ofWisconsin-Madison, Department ofMedical Physics, Madison, WI, United States ofAmerica 3 The First Hospital ofChina Medical University, Department ofNuclear Medicine, Shenyang, Liaoning, CN, People’s Republic ofChina ~~ ~~ 4 University ofLjubljana, Faculty ofMathematics and Physics, Ljubljana, SI, Slovenia Origworkinalcmayobnetentusefdromthunderis ∗ Author to whomany correspondence should be addressed. the terms ofthe Creative [E-mail:](E-mail: daniel.huff@aiq-solutions.com)[ daniel.huff@aiq-solutions.com](E-mail: daniel.huff@aiq-solutions.com) CommonsAttribution4.0 licence. Keywords: PET/CT, longitudinal, inter-reader variability, reproducibility Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. Abstract Objective. Patients with metastatic disease are followed throughout treatment with medical imaging, and accurately assessing changes ofindividual lesions is critical to properly inform clinical decisions. The goal ofthis workwas to assess the performance ofan automated lesion-matching algorithm in comparison to inter-reader variability (IRV) ofmatching lesions between scans ofmetastatic cancer patients. Approach. Forty pairs oflongitudinal PET/CT and CT scans were collected and organized into four cohorts: lung cancers, head and neck cancers, lymphomas, and advanced cancers. Cases were also divided by cancer burden: low-burden ( < 10lesions), intermediate-burden (10–29), and highburden (30+). Two nuclear medicine physicians conducted independent reviews ofeach scan-pair and manually matched lesions. Matching differences between readers were assessed to quantifythe IRV oflesion matching. The two readers met to forma consensus, which was considered a gold standard and compared against the output ofan automated lesion-matching algorithm. IRVand performance ofthe automated method were quantiﬁed using precision, recall, F1-score, and the number ofdifferences. Main results. The performance ofthe automated method did not differ signiﬁcantly from IRV for any metric in any cohort (p > 0.05, Wilcoxon paired test). In high-burden cases, the F1-score (median [range])was0.89 [0.63,1.00]between the automated method and reader consensus and0 .93 [0 .72,1 .00]between readers. In low-burden cases, F1-scores were1 .00 [0 .40,1 .00] and1.00 [0.40,1.00], for the automated method and IRV, respectively. Automated matching was signiﬁcantly more efﬁcient than either reader (p < 0.001). In high-burden cases, median matching time for the readers was60and30 min, respectively, while automated matching took a median of 3.9 min Signiﬁcance. The automated lesion-matching algorithm was successful in performing lesion matching, meeting the benchmark ofIRV. Automated lesion matching can signiﬁcantly expedite and improve the consistency oflongitudinal lesion-matching. 1. Introduction Patients with metastatic cancers are often imaged longitudinally throughout the course oftheir disease for diagnosis, staging, and response assessment with a variety ofmodalities including comp

FAQs

What was the objective of the study?

To assess how well an automated lesion-matching algorithm performs compared with inter-reader variability when matching longitudinal lesions between scans in metastatic cancer patients.

How were the scans and cohorts organized?

Forty paired longitudinal PET/CT and CT scan pairs were grouped into four cancer cohorts and further stratified by cancer burden into low, intermediate, and high groups.

What did the results show about accuracy and efficiency?

Automated matching did not differ significantly from inter-reader variability across metrics in any cohort, and it was significantly more efficient than manual reader matching.

Sep 24, 2026

Uploaded by Bintang